Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

Adventures with an ALPY and a 10 μm slit

  • Please log in to reply
35 replies to this topic

#26 robin_astro

robin_astro

    Viking 1

  • -----
  • Posts: 666
  • Joined: 18 Dec 2005

Posted 27 March 2019 - 06:40 PM

It seems to me you should have got a reasonable improvement moving to 8" aperture. The ALPY should accept light at f4 and the diffraction spot shouldn't be limiting with a 10um slit. So its reasonable to expect 75% more light and thus 75% more SNR which means you should be able to get the same SNR in a third of the exposure time (other things being equal). That's nice to have but I guess its only 0.6 in magnitude terms

 

Actually to a first approximation (ie ignoring sources of noise other than photon noise) SNR is proportional to sqrt(signal) so 1.75xsignal will increase SNR by about 32% 

 

Not sure where your 1/3 exposure time comes from. For the same signal (ignoring the effect of the central obstruction), the exposure time would be (6/8)^2 = 0.56

 

All this assumes that the star image size (a factor of both psf of the telescope optics and the seeing) is smaller than the slit (10um is very tight and needs excellent seeing)

 

Cheers

Robin


Edited by robin_astro, 27 March 2019 - 06:52 PM.


#27 robin_astro

robin_astro

    Viking 1

  • -----
  • Posts: 666
  • Joined: 18 Dec 2005

Posted 27 March 2019 - 07:08 PM

Its difficult to tell from pictures but you may benefit from a little more smoothing of the data. If you have many more than 2 pixels per optical resolution element then smoothing improves SNR without degrading spectral resolution.

 

It is a good idea to quote SNR per resolution interval (ie at 2 bins per resolution interval)to make comparisons but it is generally accepted among professionals that the niquist sampling interval does not apply to spectra. There is much debate over the optimum sampling interval but around 3-5 pixels per resolution interval is typical so smoothing spectra to lower sampling than this may risk losing information. If oversampled (unlikely in this case given the narrow slit) when using a CCD camera it is better to achieve optimum sampling by in camera binning rather than post processing filtering as you get lower read noise this way (With CMOS it makes no difference) 

 

Robin


Edited by robin_astro, 27 March 2019 - 07:09 PM.


#28 RalphC

RalphC

    Lift Off

  • -----
  • Posts: 9
  • Joined: 04 Jun 2017
  • Loc: Oxfordshire, UK

Posted 27 March 2019 - 08:29 PM

Robin,
I was just assuming camera thermal+ read noise dominate and so the signal increases and the noise doesn't (for a given exposure). I totally agree a square root SNR improvement is to be expected if photon shot noise is limiting. I was probably too hasty in making that assumption but I guess it depends what temperature the camera is. I thought colder cameras are probably better and that suggests thermal noise is significant.

#29 Organic Astrochemist

Organic Astrochemist

    Viking 1

  • *****
  • topic starter
  • Posts: 870
  • Joined: 10 Jan 2015

Posted 27 March 2019 - 09:41 PM

Hi
I've just discovered your thread which I'm reading with interest as I'm just starting out with astro-spectroscopy with some home made kit.

It seems to me you should have got a reasonable improvement moving to 8" aperture. The ALPY should accept light at f4 and the diffraction spot shouldn't be limiting with a 10um slit. So its reasonable to expect 75% more light and thus 75% more SNR which means you should be able to get the same SNR in a third of the exposure time (other things being equal). That's nice to have but I guess its only 0.6 in magnitude terms.

Its difficult to tell from pictures but you may benefit from a little more smoothing of the data. If you have many more than 2 pixels per optical resolution element then smoothing improves SNR without degrading spectral resolution.

Just for comparison on a different (but similar) system: the attached spectrum is a Wolf-Rayet hip113569 which is ~Mag 11. It was taken with my 8" scope and 2 x 320sec exposure. Of course its difficult to make direct comparisons with different slits, camera, resolution, type of star, etc but it gives an idea.

attachicon.gif hip113569 spectrum.jpg

Now I've just got to work out how to interpret spectra:)

Congratulations on your home-made spectrograph. 

Your spectrum looks great.Looks like a WN Wolf-Rayet to me.

I have found these resources helpful for spectral interpretation. There are lots more.

Spectral Atlas Book

Spectral Atlas electronic version

Astrosurf

Francois Teyssier



#30 Organic Astrochemist

Organic Astrochemist

    Viking 1

  • *****
  • topic starter
  • Posts: 870
  • Joined: 10 Jan 2015

Posted 27 March 2019 - 10:15 PM

One other bit of trivia.

One of the many things I learned at the Sacramento Mountains Spectroscopy Workshop-2 was the option in ISIS to select the Interpolator type as spline as opposed to bilinear. I don't completely understand this but Christian Buil explained that for very undersampled spectra of sharp emission lines (like in PN) the bilinear would be better. I'm much closer to the range that Robin mentioned and so I showed Christian two spectra, one processed with bilinear and the other with spline. The spline spectra revealed more fine structure. This may be contributing to the "noisier" appearance. In general, For what it's worth, the calculation of S/N with ISIS returns similar values. I don't understand enough about the relation, if any, between resolution and S/N.



#31 robin_astro

robin_astro

    Viking 1

  • -----
  • Posts: 666
  • Joined: 18 Dec 2005

Posted 29 March 2019 - 12:16 PM

Robin,
I was just assuming camera thermal+ read noise dominate and so the signal increases and the noise doesn't (for a given exposure). I totally agree a square root SNR improvement is to be expected if photon shot noise is limiting. I was probably too hasty in making that assumption but I guess it depends what temperature the camera is. I thought colder cameras are probably better and that suggests thermal noise is significant.

Hi Ralph,

 

It depends on the camera but at a typical good SNR (say 100) the photon noise normally dominates. With my cameras (ATIK 314 and 428)  the thermal noise is insignificant (effectively unmeasurable but less than one count per hour according to the manufacturer's spec). 

At low SNR on faint objects, read noise (and sky background noise) becomes increasingly important and in these cases for example it is important to minimise the number of sub exposures and  the area sampled in the image ie  keeping the spectrum narrow with good guiding and setting the binning zone tight to the spectrum. (Optimised sampling algorithms can be useful here which can automatically include only those pixels which contribute to increasing the SNR).  

 

Cheers

Robin



#32 robin_astro

robin_astro

    Viking 1

  • -----
  • Posts: 666
  • Joined: 18 Dec 2005

Posted 29 March 2019 - 12:28 PM

As a worked example,  with a perfect noise free camera and no sky noise an SNR of 100 would need 10k electrons per bin and the photon noise would be sqrt(10000) = 100 e-

 

If we consider a typical good camera with 5e-/pixel read noise and a bin size of say 2x6 pixels  and 3 sub exposures  then the total read noise/bin is sqrt(36) * 5 = 30e-

 

If we add this read noise to the photon noise we get a total noise of sqrt(100^2+30^2) =104.4  

 

The SNR with read noise therefore deteriorates to 10000/104.4 = 96,  an insignifcant change.

 

Robin



#33 robin_astro

robin_astro

    Viking 1

  • -----
  • Posts: 666
  • Joined: 18 Dec 2005

Posted 29 March 2019 - 12:32 PM

if you then run the same calculation on a noisy spectrum at an SNR of 30  (900e- /bin) 

then the total noise becomes sqrt(30^2+30^2) = 42.4

and the SNR deteriorates to 900/42.4 = 21.4,  a significant reduction

 

In this case increasing sub exposure length to reduce the number of sub exposures, binning pixels vertically (for CCD) and reducing the sampling along the spectrum to say 2-3 pixels per resolution interval can be useful 

 

(Edited 20190330 changing bin width to resolution interval)


Edited by robin_astro, 30 March 2019 - 09:09 AM.


#34 RalphC

RalphC

    Lift Off

  • -----
  • Posts: 9
  • Joined: 04 Jun 2017
  • Loc: Oxfordshire, UK

Posted 29 March 2019 - 04:12 PM

Hi Robin,
I got a little fooled by the large dark count I get on my camera. On closer examination of my dark data I see that my camera has quite an offset (~1000counts) and that gives the impression of high thermal signal but its not real. By looking at different length exposures I can see my 320sec exposures actually have only 40 counts of thermal signal per vertical bin. The photo signal in the example spectrum I gave is ~1000 counts per raw vertical bin. So you're quite right, the dark is a fraction of the photo current.

I'm still getting to know my spectroscopy camera (Its a 2nd hand Andor iVac). I don't yet know what a 'count' represents in terms of electrons (its not far off 1 count = 1 electron) but I'll figure that out soon through a series of careful noise measurements. I always worry I haven't got things optimally set up.

#35 RalphC

RalphC

    Lift Off

  • -----
  • Posts: 9
  • Joined: 04 Jun 2017
  • Loc: Oxfordshire, UK

Posted 29 March 2019 - 04:25 PM

Congratulations on your home-made spectrograph. 
Your spectrum looks great.Looks like a WN Wolf-Rayet to me.
I have found these resources helpful for spectral interpretation. There are lots more.
Spectral Atlas Book
Spectral Atlas electronic version
Astrosurf
Francois Teyssier


Thanks for the reading suggestions. They look very interesting. I really need to start understanding the spectroscopy and astrophysics more. Also Any suggestions for interesting targets gratefully received.

The hardware for my system is working ok now. Much improved now I have guiding. I still need to improve the software to automate data processing. Its a bit of a pain at present as the file format of my camera software seems incompatible with most other software. Its some weird 'fits' variant that doesn't seem readable by many packages. Ultimately I would also like to automate target acquisition (although that will have to wait a while).

#36 robin_astro

robin_astro

    Viking 1

  • -----
  • Posts: 666
  • Joined: 18 Dec 2005

Posted 29 March 2019 - 06:12 PM

Francois Teyssier's " Low resolution spectroscopy Observers's guide" from his website is an excellent source for ideas for projects using low resolution spectroscopy. (I did the rather rough and ready english translation)

 

I use Christian Buil's ISIS for all my data reduction which is pretty much the de facto standard for amateurs doing pro-am spectroscopy

http://www.astrosurf...s-software.html

It has quite a steep learning curve (though not as bad  as pro software like IRAF.) 

 

Robin


Edited by robin_astro, 29 March 2019 - 06:15 PM.



CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics